Performance Analysis of Hybrid Quantum-Classical Convolutional Neural Networks for Audio Classification

by Yash Thakar, Bhuvi Ghosh, Vishma Adeshra, Kriti Srivastava

The Research focuses on the development of a QC-CNN architecture to perform audio classification using mel-spectrograms. The QC-CNN architecture developed achieved comparative performance with classical models.

Manuscripts

This paper was presented at the 15th International Conference on Computing Communication and Networking Technologies (ICCCNT), at IIT Mandi and publised in IEEE Xplore.

DOI: 10.1109/ICCCNT61001.2024.10725668

Abstract

Audio signals being high-dimensional and complex pose challenges for classical machine learning techniques in terms of computation and generalization on real-world data. This paper evaluates the use of hybrid quantum-classical convolutional neural networks (QC-CNN) that leverage quantum effects like superposition and entanglement for audio classification using mel-spectrograms obtained from audio data. Evaluated on both small-sized and large-sized datasets, the proposed QC-CNN model gave comparable training accuracy with classical CNN (Convolutional Neural Network) on the smaller dataset but outperformed classical CNN on test accuracy (95.04% vs 92.88%) for a larger birdsong dataset and reduced overfitting, thus highlighting the potential advantages of QC-CNNs for audio data. The QC-CNN exhibited higher cross-entropy loss in case of the small-sized dataset which was further significantly reduced when evaluated on the large-sized birdsong dataset. The work demonstrates the application of QC-CNN for audio classification.

Software implementation

All source code used to generate the results and figures in the paper are in the testing folder. The calculations and figure generation are all run inside Jupyter notebooks. The data used in this study is provided in testing folder itself and the results generated by the code are plotted in plots.ipynb notebook.

Getting the code

You can download a copy of all the files in this repository by cloning the git repository:

https://github.com/yashyaks/Audio-QC-CNN.git

Dependencies

You'll need a working Python environment to run the code. The recommended way to set up your environment is through the virtualenv python package. The required dependencies are specified in the file requirements.txt.